List your company

Long-form RAG Evaluation on RAGChecker

75.6Fact RecallArbGraph
53.44859.19964.9570.701Apr 20, 2026

Evaluation results

MethodLinks
ArbGraphBackbone=LLaMA-3.1-8B, Paradigm=Ours2026.0475.699.341.24.23.636.34.651.5
ArbGraphBackbone=Qwen3-4B, Paradigm=Ours2026.0474.5135.436.64.24.933.34.453.8
MCTS-RAGBackbone=Qwen3-4B, Paradigm=Agentic2026.047298.732.6710.436.94.650.5
MCTS-RAGBackbone=LLaMA-3.1-8B, Paradigm=Agentic2026.0471.396.633.59.49.843.15.939.7
FIREBackbone=Qwen3-4B, Paradigm=Agentic2026.0470.759.634.611.68.148.67.137.8
FIREBackbone=LLaMA-3.1-8B, Paradigm=Agentic2026.0468.788.233.18.74.741.85.333.6
CRITICBackbone=LLaMA-3.1-8B, Paradigm=Agentic2026.0462.578.832.913.65.256.1731
ReActBackbone=LLaMA-3.1-8B, Paradigm=Reasoning2026.0462.162.631.512.27.7—7.125.8
CoTBackbone=LLaMA-3.1-8B, Paradigm=Reasoning2026.0459.857.830.49.46.252.56.829.8
CoTBackbone=Qwen3-4B, Paradigm=Reasoning2026.0457.6110.628.477.949.56.830.1
CRITICBackbone=Qwen3-4B, Paradigm=Agentic2026.0454.841.8254.85.2456.935
Standard RAGBackbone=LLaMA-3.1-8B, Paradigm=Naive2026.0454.751.729.712.5659.96.929.2
ReActBackbone=Qwen3-4B, Paradigm=Reasoning2026.0454.44528.97.19.455.26.535.1
Standard RAGBackbone=Qwen3-4B, Paradigm=Naive2026.0454.375.831.57.97.751.46.632.2